Predicting in-hospital outcome in HIV-associated Pneumocystis carinii pneumonia

Abstract

Pneumocystis carinii pneumonia (PCP) in HIV-infected patients remains a life-threatening complication in the course of HIV infection. Despite effective treatment, mortality may still be as high as 10%. The identification of risk factors associated with a lethal outcome might be helpful as a guide to therapy for patients at risk and in the evaluation of new drugs with anti-pneumocystic activity. In a retrospective study 58 first episodes of HIV-associated PCP without prophylaxis were analyzed. Variables associated univariately with higher mortality were identified. A prognostic rule was established in a multivariate approach using canonical discriminant analysis. Cut-off values for parameters included were determined in order to allow a clinically applicable estimate of the individual risk. Variables associated with early mortality were hemoglobin, hematocrit, platelet count, albumin, total protein, gamma-globulins, and AaDO2. LDH values, percentage of neutrophils in the BAL, as well as the cellular immunologic state as indicated by CD4-cell count were not significantly associated with the outcome. The discriminant function yielded the best classification results with the inclusion of hemoglobin, albumin, and gamma-globulins (overall accuracy 86%). Two or more of the following parameters were associated with a 14-fold increased risk of in-hospital mortality: hemoglobin less than 10 g/dl, albumin less than 3 g/dl, and gamma-globulins less than 1.2 g/dl. This prognostic rule was 80% sensitive and 94% specific with a negative predictive value of 94%, yielding an overall accuracy of 91%. Patients with HIV-associated PCP with a positive prognostic rule have a 14-fold increased risk for in-hospital lethal outcome. This discriminant rule may be helpful in identifying patients at risk.